Selection criteria

The following inclusion criteria, based on the PICO (Population, Intervention/Exposure, Comparison, Outcomes) model were applied [16]:

• Population: pregnant women, health professionals, and infants aged 0-5-years-old, including their parents. This age range was chosen for the inclusion of many of the early routine vaccinations administered across vaccination schedules of most countries [17].
• Exposure: defined as the COVID-19 pandemic, as declared by the WHO on 11^th March 2020 [18].
• Comparison: defined as the pre-COVID-19 pandemic period, any period prior to March 2020 where the WHO declared a global pandemic, this has also been defined by the studies included themselves [19].
• Outcomes: changes in routine maternal and infant vaccination coverage, vaccination services (for example, operating hours, changes in delivery schedules), and/or vaccine confidence.

The WHO definition of routine vaccination (as stated above) was used [3]. Quantitative and mixed methods studies were included to gather all relevant quantitative results, and all countries were included for a global perspective. Studies were excluded if:

• They were not presented using English language to avoid translation error,
• They focused on other coronaviruses, for example SARS-CoV-1.
• The study PICO differed from those specified above.
• The sole focus was on non-routine vaccinations administration, considered as vaccinations not found on routine vaccination schedules such as post exposure prophylaxis, including the recent COVID-19 vaccine.

Search strategy

A search strategy was created using relevant medical subject headings (MeSH) [20] and free text search terms, examples include pregnant, infant, vaccination, and COVID-19 (see S1 Text), to identify suitable studies. Databases and sources searched include OVID Medline (R and Epub Ahead of Print, In-Process, In-Data-Review and Other Non-Indexed Citations and Daily), EMBASE, CINHAL, PsychINFO, Web of Science (Social Science Citation Index), Google Scholar, WHO, UK Government JCVI (including EU and US equivalents), and SAGE Journals. Pre-prints were included within searches to gather all relevant data. Searches occurred between 15-17th June 2021, and publications up to the search date were included (excluding all studies published pre-2020 due to irrelevance to COVID-19). Retrieved studies were uploaded to the reference management tool EndNote. Duplicate studies were removed, and remaining studies were screened using their titles and abstracts to decide upon their relevancy to the review; this process was carried out only by AY due to resource constraints. Decisions were recorded using a PRISMA flow diagram [21]. Full texts of relevant studies were retrieved for full eligibility checks following title and abstract screening. Decisions around inclusion of studies where eligibility was less clear were made via team discussion (all authors) by strictly comparing these studies to our pre-defined PICO and considering resource constraints in the potential widening of this PICO for the inclusion of these studies.

Data extraction

The author, year of publication, country, country income (based on World Bank 2021 classification) [22], study purpose, data collection methods and sources, population, sample size, exposure, control, outcomes, and other data of importance were extracted on to a data extraction form using Microsoft Excel by one researcher (MSc AY) due to resource constraints (see Table A in S1 Table). This enabled the comparison of differences between studies. Summary estimates, including confidence intervals, and p-values of quantitative studies were extracted where possible for the comparison of effect estimates between pre- and post-COVID-19 periods. Only quantitative data were extracted from mixed method studies.

Quality assessment

A Risk of Bias (ROB) assessment, using the Newcastle-Ottawa Scale (NOS) [23], was applied to all included studies, as recommended by the Cochrane handbook [15]. The NOS adapted for cohort studies was applied to one study which specified itself as a cohort study. This was scored out of a maximum of 9 stars [23]. The NOS adapted for cross-sectional studies was applied to all other studies as they either defined themselves as cross-sectional or were not explicitly clear on their study type but could be identified as cross-sectional studies [24]. These studies were scored out of a maximum of 10 stars [24]. On the NOS scale, a score of ten stars represents low ROB while zero stars represents very high ROB [23]. The NOS simultaneously acted as a quality appraisal tool [23].

Data synthesis

A narrative synthesis was conducted focusing on vaccine coverage, vaccination services, and vaccine confidence as outcomes. This was appropriate due to the variation between studies in their chosen methods of reporting (cumulative counts vs rates), while also allowing cohesive discussion of interactions between the different outcome measures stated [25, 26]. Tabulation of data was used throughout to assist with the presentation of results and to enable comparison between HICs and LMICs. Due to heterogeneity between studies and results, a meta-analysis could not be conducted [15]. The influence of ROB on the results of the review were explored.

Study characteristics

Study characteristics are presented in Table 1 along with the main study results. Two studies were classified as descriptive analysis studies [32, 51], one as an interim analysis [35], eight as observational studies [27] (including ecological [37], cross-sectional [33, 39, 53, 54, 56], and cohort studies [31]), two as mixed method studies (one descriptive analysis [36], one cross-sectional [43]), and one as a retrospective review [48]. Sixteen studies did not specify their study type, and so were categorised as observational studies [28–30, 34, 38, 40–42, 44–47, 49, 50, 52, 55] (see Table 1). Twenty studies focused on HICs (Japan [27], US [28, 29, 32, 38, 40–42, 44], Netherlands [30], Singapore [31], Canada [33], England [34–36, 43], South Korea [37], Sweden [39], Italy [45, 46]), seven focused on LMICs (Pakistan [47, 50], South Africa [48], Brazil [49], Nigeria [51], Turkey [52], Lebanon [53]), and three studies focused on larger geographical areas including multiple countries which will be referred to as regional studies (global [56], Africa [55], South-East Asia and Western Pacific [54]), listed in Table 1. One study analysed multicentre data from a number of vaccination service providers (9 healthcare facilities across Singapore) [31], 19 studies analysed routinely collected data (eight at national-level [30, 34, 35, 37, 39, 41, 49, 55], and 11 at regional-level within the country [27–29, 32, 38, 40, 47, 48, 50–52]). Ten studies reported project-level data by independently gathering participants and collecting data through means such as surveys, questionnaires, and individual and service provider records [33, 36, 39, 42, 43, 45, 46, 53, 54, 56].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of studies included in the systematic review.

https://doi.org/10.1371/journal.pgph.0000628.t001

Two studies researched routine vaccinations for pregnant mothers [34, 50], and 29 for infants aged 0-5-years-old [27–33, 35–56]. Fifteen study populations also included those above 5-years-old: data were only extracted for vaccinations administered to those aged 0-5-years-old from these studies [27–29, 36, 37, 40–42, 44–46, 51, 53, 54, 56] (Table 1). Nine studies lacked a defined pre-COVID-19 period, however, were included as they reported data on the impacts of the pandemic and made it clear data comparisons were made to general pre-COVID-19 periods [33, 36, 39, 43–46, 54, 56].

Vaccinations in the studies include:

• Mumps, Measles, and Rubella (MMR)
• MMR and Varicella (MMRV)
• Measles containing vaccine (MCV)
• Diphtheria, Tetanus, and acellular Pertussis (DTaP)
• Diphtheria, Tetanus, Pertussis, Polio, Haemophilus influenzae type b (5-in-1)
• 6-in-1 (equivalent of the 5-in-1 and Hepatitis B)
• Pneumococcal conjugate (PCV)
• Bacillus Calmette-Guérin (BCG)
• Hepatitis B (HepB and 1^st dose HBV0)
• Haemophilus influenzae type b (Hib)
• Hepatitis A (Hep A)
• Influenza
• Polio (including Oral Polio vaccine OPV, Inactivated Polio Vaccine IPV)
• Japanese Encephalitis (JE)
• Varicella
• Rotavirus (Rota-1).

Risk of Bias

18 studies achieved a NOS score of 7 stars or above, and therefore could be considered as good studies with a low ROB [27–37, 47–51, 54, 55] (see Tables B and C in S1 Table). 12 studies obtained a score less than 7 stars indicating increased ROB due to: no statistical tests, poor comparability by disregarding relevant confounders (including the age of infants at time of vaccination, or service type as public or private), or sampling concerns (small sample size, or convenience sampling) reducing the representativeness of the study population [38–46, 52, 53, 56]. Studies are presented in Table 1 and Tables A-C in S1 Table in descending order of ROB score.

Vaccine coverage

Results have shown an overwhelming decrease in vaccine coverage. 18 studies examined infant vaccination coverage, representing all seven LMICs [47–53], nine HICs [28, 30, 32, 34, 35, 37–39, 46] and two larger regional studies [54, 55]. 17 of these studies (13 low ROB [28, 30, 32, 34, 35, 37, 47–51, 54, 55], 4 high ROB [39, 46, 52, 53]), reported decreases in vaccination coverage including for varicella, JE, PCV, HepA, BCG, HepB/HBV0, DTaP, Polio/OPV/IPV, and MMR/MMRV/MCV, though four of these also reported some mixed results [32, 37, 38, 55]. One study (with a higher ROB) reported only an increase in vaccination uptake (first dose MMR, US) [39] (see Table 1). Between all studies included in this review decreases up to -79% were seen across all vaccinations [28, 30, 32, 34, 35, 37, 38, 46–55].

We compared our most robust studies (ROB score ≥ 7) that reported changes in vaccination coverage between HICs and LMICs. Five of the six robust HIC studies of vaccination coverage showed a decline ranging from -1.8% (MMR1 schedule completion in infants by 6 months old, national immunisation data, England [35]) to -14% (MMR1 in infants < 2 years, national immunisation data, Netherlands [30]), with others in between (e.g., -5% decrease in fully immunised 24-month-olds, regional immunisation data, Texas, US [32]). One study showed slight increases of 0.3–1.4% across specific infant vaccinations in South Korea [37]. Five of the most robust LMIC studies showed a decline ranging from -3.7% (BCG coverage in < 5-year-olds, regional data, Oyo State, Nigeria [51]) to -24% (BCG coverage, Brazil [49]) decrease, with others in between (e.g., -7% decrease in fully immunised infants <1-years-old, regional data, KwaZulu-Natal Province, South Africa [48]). The large African regional study reported mixed results across countries for DTP3 and MCV1 vaccination coverage, including some decreases (up to -52% for DTP3, Guinea) and some increases (up to 13% increase for MCV1, Chad) [55]. Increases were explained by the authors as resulting from a lack of strict, extended lock-down periods, and COVID-19 cases [55]. See Table 1 for more details.

Harris et al.’s (low ROB) large regional study included both HICs and LMICs and reported an overall decline for DTP, OPV, IPV, and Measles vaccine coverage rates within all ages; the greatest being in OPV with a median decrease of -79% (IQR -42% to -79%) administered during infancy in participants from 19 different countries across South-East Asia and the Western Pacific [54]. The smallest decrease was reported within school-entry aged children receiving measles vaccination with a median decrease of -9% (IQR -3% to -31%), from the same study [54].

The two studies exploring maternal vaccination coverage, both reported decreases. Chandir et al., reported a -28.8% average decrease in maternal tetanus toxoid vaccinations in (LMIC) Pakistan, while Public Health England reported a -4.2% decrease in monthly maternal pertussis vaccination coverage in (HIC) England [34, 50].

Studies with a high ROB (ROB score < 7), show more conflicting findings in coverage, though still mainly indicating a decline.

Vaccine service changes

Post-pandemic results show a decrease in vaccine administration and disruptions to services in comparison to the pre-pandemic period, as reported in 21 studies: two regional studies [54, 56], six LMICs [47, 48, 50–53], and 13 HICs [27, 29, 31–33, 36, 39–41, 43–46] (Table 1). 11 low ROB [27, 29, 31–33, 36, 47, 48, 50, 51, 54] and 10 high ROB [39–41, 43–46, 52, 53, 56] studies exploring changes in vaccination services stated a negative effect direction, representing decreases in administrations and difficulties in vaccination delivery and access. Aizawa et al., reported conflicting results, due to differences in the administration of vaccines between age groups; increases in 5-6-year-olds receiving the MR2 vaccine [27]. Examples of these disruptions follow below including the extent of how these differ based on the characteristics of the vaccine services (e.g., public or private sector) or the infants (e.g., age).

Results showed decreases in vaccine administrations from the pre-pandemic period; for example, a -15% to -7.5% decrease in BCG administration in Japan [27]. Some changes to vaccine schedules were seen such as the mean age of BCG vaccination administration decreased from 6.3-weeks prior to the lockdown, to 4.3-weeks-old (95% CI 1.93 to 2.07, p < 0.01) in Sindh, Pakistan [50]. Some service providers only continued vaccinations for certain ages; for example, Vogt et al., found 81.4% of services in the US offered vaccinations to 1–2-year-olds, whereas only 44% continued for 3–6-year-olds [44]. Likewise, Piché-Renaud et al., identified 94% of services in Ontario, Canada continued vaccinations for 0-18-month-olds, while 77% postponed vaccinations for 4–6-year-olds [33]. Overall, declines in vaccine administrations were reported across both LMICs and HICs. These were more common within LMICs as in some cases vaccination administrations increased in HICs. For example, a 2% to 7% increase in measles and rubella 2^nd dose (MR2) vaccine administrations for infants aged 5–6-year-olds across within-country regions in Japan (Kawasaki, Niigata, Nagasaki, and Fuchu) [27].

Declines in administrations were greater within private sectors [28, 31, 47, 50, 53, 54]. In Singapore the number of MMR/MMRV vaccines administered differed between polyclinics -25.6% (95% CI -28.1% to -43.3%), hospitals -57.3% (95% CI -65% to -50%), and private clinics -73.6% (95% CI -81.0% to -65.1%) [31]. Harris et al., reported 79% of public sector antigens (vaccinations) were disrupted, 83% within the private sector within the South-East Asian and Western Pacific region [54]. Decreases in vaccine administrations between public and private sectors were primarily seen within HICs, whereas in LMICs these differences in vaccine administrations by setting were typically reported between fixed and outreach services. Declines in administrations were greater for outreach services (-79.3% to -88.6% decrease) than for fixed-centre services in Pakistan (-32.1% to -38.7%) [47, 50]. Interestingly Olaniyan et al. reported a 3.1% increase in fixed and 4.2% increase in outreach vaccination services within Oyo State Nigeria, however, a decrease in vaccination coverage (-3.5%) for HBV0, could still be seen in this study [51].

Locations of vaccination services impacted the extent of service delivery in both LMICs [48, 50] and HICs [32, 44]. In Texas US, vaccine administration for 5-month-olds declined by -28.6% (95% CI -21.7% to -25.7%) in rural areas compared to -1.4% (95% CI -1.7% to -1.2%) in urban areas, while for 16-month-olds vaccination declined by -12.1% (95% CI -15.5 to -8.5) in rural areas, compared to -18.2% (95% CI -19.1% to -17.3%) in urban areas [32]. eTheKwini, an urban area in South Africa, reported a -37% decline in measles 1st dose coverage, whereas rural Zululand experienced a -10% decline [48].

Results show reductions in operating hours and increased duration of consultations [33, 36, 39, 43, 44, 46, 50, 51, 54, 56]. For example, Vogt et al., found across the US, 61.7% of practices offered reduced office hours for in-person visitations; of these, 63.7% were in urban areas, and 55.4% in rural areas [44]. Across other studies, Russo et al., found up to 42.5% of vaccination appointments were postponed or cancelled by vaccination services from their 1,474 survey responders in Italy, 13.5% stated vaccination services closed, while 44% of parents were reluctant to travel due to travel restrictions [46]. A lack of guidance was identified by England as Bell et al.’s online survey which found 25.6% of parents were unaware that childhood vaccinations continued throughout the pandemic [43]. From this same study, 23.9% to 53.3% of parents experienced difficulties accessing and booking their child’s vaccination appointment [43]. Logistical disruptions included: staff shortages, for example as identified by Saso et al. in their globally distributed questionnaire; equipment shortages, including personal protective equipment (PPE), and issues with the vaccine supply-chain [33, 36, 39, 54, 56]. Sindh, Pakistan experienced a -7.4% (95% CI -5.29% to -9.51%, p < 0.0001) decrease in the daily average vaccinator attendance, a common occurrence in LMICs [50].

Vaccine confidence

Six studies addressed differences in vaccine confidence between the pre- and post-pandemic period; terminology included ‘vaccination intentions’, ‘importance of vaccination’, and ‘parent concerns’ [33, 39, 42, 43, 45, 52].

Some differences in vaccine confidence were reported among parents. Sokol and Grummon found 60% of parents intended to change their paediatric influenza vaccination behaviour due to the pandemic [42]. For parents whose children did not receive the 2019–2020 influenza vaccine 34% (95% CI 30%– 27%) responded that the pandemic made them less likely to have their child vaccinated for the 2020–2021 influenza vaccine compared with their plans before the pandemic, while 21% (95% CI 18% - 24%) responded they would be more likely [42]. Among parents whose children received the 2019–2020 influenza vaccine, 24% (95% CI 22% - 27%, p < 0.001) reported being less likely, while 38% (95% CI 35% - 41%) reported being more likely to have their children vaccinated with the 2020–2021 influenza vaccine [42].

Between practitioners, in Turkey, when asked ‘Which was the attitude of your patients regarding routine vaccination during the pandemic?’ 38.3% of family practitioners, 74.4% of paediatricians, and 65.8% of paediatric infectious disease specialists stated patients did not want to come in for vaccination due to the pandemic [52]. However, 57%, 10.5%, and 11.4% respectively also stated no problems with parental attitudes during the pandemic [52]. In Sweden, physicians reported parental concerns over their infant’s vaccination administration by comparing the post-pandemic attitude of parents to the pre-pandemic period, on a scale of 1 (not at all concerned) to 10 (very much concerned); 5% reported a score of 5, 10% a score of 4, 15% a score of 3, 40% a score of 2, and 30% a score of 1, signifying a fair proportion of parents with increasing concerns surrounding the vaccination of their infant following the pandemic [39].

Strengths and limitations

A strength of this review is the comprehensive investigation into an important health area impacted by the pandemic with potentially significant public health consequences. By conducting a quality assessment and comparing the outcomes of high and low ROB studies we were able to strengthen our conclusions.

Limitations include that due to resource constraints, one researcher conducted screenings and data extraction. While the methods would be strengthened by independent screening and data extraction by another researcher, cases of uncertainty were discussed in depth with two experienced researchers (co-authors EA and CC) to minimise this limitation. Due to time constraints, qualitative studies were not explored, which we recognise as beneficial to include in future research to provide richer detail on these findings. We identified a lack of research on maternal vaccinations so could not draw strong conclusions about the pandemic effects, though the existing research indicates cause for concern. We recommend more research be done in this area with the inclusion of qualitative studies for a richer explanation of results. Four studies identified during the literature search were not published for public use, while two studies were not presented using English language, resulting in potential missing evidence. Heterogeneity between studies prohibited the modelling of a comprehensive meta-analysis. In the future a review investigating the impacts of the pandemic between specific time periods, for example pre-lockdown vs lock-down periods, may assist in understanding the extent of the impacts of the pandemic.