Vaccination coverage at seven months of age in Limpopo Province, South Africa: a cross-sectional survey

Many low- and middle-income countries face challenges in attaining adequate levels of vaccination coverage, and the factors driving this under-coverage have not been completely elucidated. In this cross-sectional study, we investigated factors associated with vaccination coverage in Mopani District, Limpopo Province, South Africa. Between July and October 2017, we surveyed 317 caregivers (83% of whom were mothers) of seven-month-old infants in Mopani District about barriers faced when attaining vaccines and attitudes towards vaccination, and reviewed the infants’ documented vaccination history. Caregiver and child demographic data were collected shortly after birth. We described the coverage for vaccines that should be received by age seven months, according to South Africa’s Expanded Programme on Immunization schedule, and explored the relationship between coverage and caregiver characteristics, behavioral factors (e.g. attitudes towards vaccination), and structural factors (e.g. vaccination stock-outs at clinics). We found that caregivers reported positive attitudes towards vaccination, based on a seven-question survey of vaccination attitudes. Although coverage was high for most recommended vaccines, it was low for pneumococcal conjugate vaccine (PCV), with just 36% of children having received it by age seven months. This appears to have been due to PCV stock-outs at government clinics. For vaccines other than PCV, children were more likely to be up-to-date on vaccinations if a community health worker (CHW) had visited their home in the past month (adjusted odds ratio (OR) 1.24, confidence interval (CI) (1.10–1.41); p < 0.001) and if the caregiver had more years of schooling (adjusted OR 1.03 (CI 1.01–1.05); p = 0.012). We conclude that addressing PCV stock-outs at government clinics in Mopani District is necessary to ensure coverage reaches adequate levels. Additionally, supporting CHW programs may be a productive avenue for improving vaccination coverage.


Introduction
The expansion of routine childhood vaccinations has led to significant reductions in childhood morbidity and mortality globally. The World Health Organization (WHO) estimates that vaccines for just four diseases -diphtheria, tetanus, pertussis, and measles -prevent two million to three million deaths each year. However, coverage rates remain low in some settings, and there are still 1.5 million deaths annually from vaccine-preventable diseases (1). While there have been large studies on attitudes towards vaccination in sub-Saharan Africa, such as a recent survey of attitudes towards COVID-19 vaccination in 15 African Union countries, there has been less study on the vaccination attitudes of caregivers of infants in sub-Saharan Africa (2). This is problematic because South Africa has had issues with childhood vaccination coverage. At the time this study was conducted, for a variety of routine childhood immunizations, the WHO and United Nations Children's Fund (UNICEF) estimated only 66-78% of South African children are vaccinated on the recommended schedule, although the South African government's coverage estimates were higher for each vaccine (84-96%) (3).
Knowledge and attitudes about vaccines among new or expecting mothers have been studied in some sub-Saharan African countries (e.g. Uganda, Nigeria), with some evidence that mothers with more positive views of vaccines are more likely to have their children vaccinated (4,5). There is a need for further study of how new mothers' knowledge and attitudes may contribute to vaccination coverage in South Africa specifically. A study of nurses providing immunizations in Limpopo Province identified caregiver non-compliance with the recommended vaccination schedule as a major contributor to poor vaccination coverage, although it was not determined how much of this noncompliance was due to limited knowledge, negative attitudes, or structural factors (e.g. long wait times at clinics) (6).
In addition to caregivers' knowledge of and attitudes towards vaccines, structural factors may affect vaccine coverage in South Africa. For instance, community health workers (CHWs) have been shown to increase vaccination coverage in other parts of sub-Saharan Africa (7). Conversely, barriers to vaccination reported by caregivers in other parts of sub-Saharan Africa include distance to clinic, hours of service at clinic, and long queues at the clinic (8)(9)(10). In South Africa, people receive vaccinations at government primary care clinics for free, thus the direct cost of the vaccination should not be a barrier to caregivers' having their children vaccinated. However, stock-outs (i.e. shortages of at least one vaccine on the day of contact) would prevent parents from having their children receive that vaccine.
In this study, we evaluated some of the behavioral and structural factors which have influenced vaccine coverage among a cohort of 317 caregiver-child dyads in Limpopo Province, South Africa.

Setting
This study took place in the town of Tzaneen in Mopani District, Limpopo Province, South Africa. Mopani has a population of 1.1 million people, around 80% of whom live in rural areas (11). Tzaneen is located 180 km from the Zimbabwe border, 250 km from the Botswana border and 170 km from the Mozambique border. The Greater Tzaneen municipality has a population of 390,095, with a population density of 120 persons/km 2 (12). The two largest ethnolinguistic groups in Tzaneen are speakers of Sepedi (Northern Sotho) and Xitsonga.

Sampling
This cross-sectional study was nested within a larger, two-year-long child development intervention trial (SANCTR: reference number 4407/ PACTR201710002683810) (13). In this trial, a package of CHW interventions, largely centered on promoting positive mother-child interactions, is being tested to see what effects they may have on children's physical and neurocognitive development during the first two years of life. Overall, 1107 caregiver-child dyads have been enrolled in the trial and are being followed for two years. Children were enrolled from birth and randomized to either intervention (i.e. curriculum integrated into existing CHW protocol with monthly visits to the household) or control arm (i.e. standard protocol). At enrollment, information on household demographics, child health, and caregiver well-being were collected through face-to-face interviews. At endline (two years of age), the impact of the intervention on child health and physical and neurocognitive development will be assessed. A subsample of 317 caregiver-child dyads is participating in a sub-study. Participants in the substudy visit a testing center in the town of Tzaneen for neurocognitive measurements, principally eyetracking and electroencephalography, when children are aged seven, 16, and 25 months. A cross-sectional survey on vaccine knowledge and attitudes was administered to caregivers when children visited the testing center for their seven-month assessment. The other data presented in this paper are secondary data from the larger child development study. For our study, 'caregiver' was defined as the adult family member who brought the child in for testing, which in this study tended to be mothers (Table 1).

Data collection
Data collection forms were used to gather demographic information from the caregivers, as well as the children's vaccination data. Caregiver and child demographic information was collected at the time of enrollment in the study. The vaccination data were acquired from the Road to Health Card (distributed by the National Department of Health) that caregivers were asked to bring to the testing center. This card lists the age-appropriate vaccinations that children should receive according to recommendations from the South African Expanded Programme on Immunisation (EPI) (14). Health care workers record the date the vaccination was given and the batch number in the card. The survey vaccine questions were based largely on questions developed by World Health Organization's (WHO) Strategic Advisory Group of Experts on Immunization (SAGE) Working Group on Vaccine Hesitancy, as well as questions from the Parent Attitudes about Childhood Vaccines survey (15)(16)(17). The survey questions were written in English and then verbally translated into the local languages, Xitsonga and Sepedi, by local research staff, who administered the survey to the study participants. The study team agreed on appropriate wording for the translations during the training and piloting phase. The results of these surveys, as well as general demographic data, were entered into an electronic REDCap database (Vanderbilt University, Nashville, USA) by local research staff (18).
Ethical clearance for the main intervention trial, in which this survey was nested, was obtained from the institutional review boards of Boston University, University of the Witwatersrand, and the Limpopo Department of Health. Written informed consent for study participation was obtained from caregivers in the local languages, Sepedi and Xitsonga, with the help of local research staff who served as interpreters.

Variables
To measure caregivers' views on vaccines, a simple vaccine positivity index was created. During the survey, caregivers were asked to agree or disagree with seven different statements concerning vaccinations, with the options 'Strongly agree', 'Agree', 'Not sure', 'Disagree', or 'Strongly disagree.' For the five statements in which agreement indicated a more positive view about vaccination, 'Strongly agree' and 'Agree' were both coded 1. For the two statements in which agreement indicated a more negative view about vaccination, 'Strongly agree' and 'Agree' were both coded −1. The caregivers' responses to the seven statements were summed to create a vaccination positivity index, which could range from a low of −2 (uniformly negative views about vaccination) to 5 (uniformly positive views about vaccination). A continuous wealth index was constructed based on principal component analysis of household asset data per the methodology described by Filmer and Pritchett (19). Participants reported whether their household owned 27 key assets (e.g. a radio, a television, a refrigerator, etc.), selected based on the questions from the 2014/2015 South Africa Living Conditions Survey (20).

Analysis
All data were analyzed using R (R Foundation for Statistical Computing, Vienna, Austria). The primary outcome of interest was 'full coverage,' or whether a child had received their recommended vaccinations by seven months of age. 'Full coverage' was defined for this study as having received Bacillus Calmette-Guérin tuberculosis vaccine at birth; oral polio vaccine at birth and six weeks of age; diphtheria and tetanus toxoids, acellular pertussis adsorbed, inactivated poliovirus, Haemophilus influenzae type B, and hepatitis B vaccines at six, 10, and 14 weeks; and rotavirus vaccine (RV) at six weeks. National guidelines recommends that by six months of age, infants should have received all of these vaccinations at the timeframes indicated, as well as pneumococcal conjugate vaccine (PCV) at six weeks and 14 weeks, RV at 14 weeks, and measles vaccine at six months (14). We excluded PCV in our definition of 'full coverage' because it appears that PCV was not available during the study period due to stock-outs, a factor outside of the control of caregivers whose inclusion in our model would make it difficult to elucidate any other factors correlated with immunization status. Measles vaccine was also excluded because the recommended date of administration (six months) and the seven-month assessment visit were so close in time, and did not give caregivers adequate time to present at the clinic and receive the vaccination before assessment. Additionally, RV is recommended at both six weeks and 14 weeks, but the 14-week vaccination was not asked about in our survey, so was necessarily excluded from our definition.
Demographic characteristics are summarized using frequencies for categorical variables and means with standard deviation for continuous variables. We present the proportion of children who had received each vaccination at any point before seven months of age and not necessarily by the time recommended by the government (e.g. a child who at age three months had received a vaccine recommended at one month would be counted as having coverage of this vaccine). We also used logistic regression to identify factors associated with 'full coverage' by age seven months. Table 1 shows demographic information for the caregivers at the time of study enrollment. The majority (88%) of the caregivers who brought children to the seven-month assessment were the same as those who completed the baseline enrollment survey. Most caregivers (83%) who brought the child to their seven-month assessment were the child's mother. The average age of caregivers in this study is 31.8 years old (SD 10.4 years), with an average of 10.2 years of schooling completed (SD 2.6 years). Twelve percent of caregivers were employed outside of the home and 42% were married. Additionally, 31% of caregivers surveyed were first-time mothers. While just 29% of caregivers reported clean water on site at home, over 99% of households reported owning a cell phone. Many caregivers (45%) reported that a CHW had visited in the past month, and 38% reported that the CHW had spoken with them about vaccines at their most recent visit.

Results
The vaccination status of children enrolled in the study was based on Road to Health Cards, which all caregivers were asked to bring to each study visit and which 99% of caregivers brought to their sevenmonth visit. Of note, while vaccination coverage was generally high in this population, coverage was lower for measles (87%) and very low for PCV (36%) ( Table 2). In total, only 29% of children were up-to-date on all vaccinations they were supposed to have received by age six months. However, coverage of all vaccines besides PCV was 70%. Furthermore, looking at coverage for all vaccines besides both PCV and measles (i.e. 'full coverage'), this rose to 76%. Notably, coverage for individual vaccinations in the study population was higher than the most recent WHO and UNICEF estimates for South Africa as a whole at this time of this study (3), as seen in the final column of Table 2.
Overall, for the vaccination positivity index, which ranges from −2 (most negative) to 5 (most positive), the mean was 4.1 (SD = 1.1), indicating that caregivers generally had positive feelings about vaccines (Figure 1). For instance, 99% agreed it is important for themselves and their children to get recommended vaccinations. Of note, though, a number of caregivers did not feel confident the clinic would have the vaccinations they require when they need them (23%). Additionally, one in five caregivers felt they were not provided with enough information about vaccines, their safety, and how they work (20%). While several caregivers indicated that barriers for accessing childhood vaccinations included the distance from their home to the clinic (12%) and time needed to reach and wait at the clinic (11%), by far the most commonly cited barrier was that the vaccine was out-of-stock at the clinic (50%) (Figure 2). A higher proportion of caregivers whose children did not receive PCV reported vaccines were out-of-stock compared with caregivers whose children did receive PCV (53% and 41%, respectively; p = 0.026). Thus, it appears stock-outs are a major factor driving low PCV coverage. Of note, none of the caregivers indicated vaccine cost was a barrier for them. Table 3 shows the results of a logistic regression model exploring which factors contributed to whether or not a child achieved 'full coverage' of recommended vaccinations by seven months of age. There was a significant positive association between 'full coverage' and the dummy variable of whether a CHW had visited in the past month (adjusted odds ratio (OR) 1.24 (1.10-1.41), p < 0.001). Additionally, there was a smaller but still significant positive association between 'full coverage' and caregiver education, in years of schooling (adjusted OR 1.03 (1.01-1.05), p = 0.012). Vaccine positivity index of caregiver, total barriers reported by caregiver, CHW discussion of vaccines at most recent visit, wealth index of household, caregiver age, caregiver parity, and caregiver employment were all found to be statistically insignificant in relation to whether a child had achieved 'full    In total, 206 caregivers (65.0%) reported at least one barrier to vaccination. A mean of 0.8 barriers to vaccination was reported (SD 0.7). coverage.' There were 37 participants excluded from the logistic regression model due to missing values for one or more variables included in the model. Participants dropped from the final analysis due to missing data were not systematically different from those who were included (Appendix 1).

Discussion
This analysis yielded four main findings. First, caregivers generally expressed positive attitudes about vaccination. Second, vaccination coverage rates were high except for PCV. Third, low PCV coverage appeared to be related to stock-outs at government clinics. Fourth, recent CHW visits and caregiver education were both associated with higher probability of achieving 'full coverage' by seven months of age.
Attitudes towards vaccinations were generally positive among the caregivers in the study population, but it appears structural factors determined vaccination coverage more than behavioral factors (e.g. knowledge and beliefs about vaccines). In particular, vaccine stockouts were a common problem among this population, particularly for PCV. South Africa first introduced PCV to its recommended EPI vaccine schedule in 2008, in the form of the older formulation PCV7, before introducing the newer formulation PCV13 in 2011 (14). The introduction of PCV dramatically reduced rates of invasive pneumococcal disease among South African children younger than two years old (21). It is currently recommended that infants receive three different PCV shots, at six weeks, 14 weeks, and nine months of age. However, when PCV was first introduced in South Africa, there were reports of widespread stockouts, with several reports of continued PCV13 stockouts in South African government clinics since then. For instance, one study in the Tshwane Health Province of Gauteng Province found 65% of surveyed clinics had experienced a stock-out of PCV13 in the previous 12 months, with similarly high stock-out rates for other vaccines like rotavirus and DTaP-IPV/Hib (22). In Limpopo Province specifically, the Stop Stock Outs Project has uncovered notable vaccine stock-outs; in 2014, for example, 28% of clinics reported they did not have any DTaP-IPV/Hib vaccine in stock (23). For PCV13 in particular, high prices pose a major barrier to access in South Africa, since the country spends more than 30% of its annual vaccination budget on PCV13 alone (24).
The most statistically significant finding of the regression model was a positive association between 'full coverage' and a CHW visit in the past month. This appears to support previous findings which link CHW programs in South Africa with higher rates of vaccination coverage (25). Additionally, there was a significant positive association between 'full coverage' and the number of years of schooling completed by the caregiver. This appears to support previous research which found maternal education has a positive association with child vaccination (26). It is plausible that caregivers with a higher level of education were better able to understand the instructions given to them about when to bring their children in for vaccination, although further study is needed to better understand the mechanism behind this association. It is interesting to note that, although not statistically significant, CHW discussion of vaccines with caregiver at last visit was associated with a decreased likelihood of 'full coverage.' It is plausible this is because CHWs were more likely to discuss vaccines with caregivers whose children were not up-to-date on vaccinations; however, given the lack of statistical significance, further study is needed on this topic.
There are several limitations of this study. Surveyed caregivers were all enrolled in a larger child development study and thus may not be representative of the full population of caregivers in Limpopo Province, because caregivers who participated in this study may be better integrated into the formal health system than other caregivers in this area. Another limitation is that this study only gathered data on whether children had received their first RV and PCV at six weeks of age, even though by six months of age they should all have received two RV and two PCV (at six weeks and 14 weeks for each vaccine). Therefore, it is difficult to accurately assess complete coverage of the PCV and rotavirus vaccine series because coverage of the second vaccine in each series (at 14 weeks) was not recorded. Additionally, the generally positive attitudes about vaccines expressed by caregivers may in part be due to a social desirability bias, whereby respondents expressed these positive views about vaccines because they believed it was what surveyors wanted to hear. Furthermore, the possible barriers to vaccination which we investigated were not exhaustive, and many other conceivable barriers (e.g. immigration status, religious beliefs, behaviors of local healthcare workers, etc.) were unexplored and may warrant further investigation. Finally, because of the descriptive nature of this crosssectional study, it is not possible to demonstrate causation from the associations described. Therefore, any conclusions from this study must be treated with caution and further study is warranted in order to better understand the relationships between vaccination coverage and stock-outs, CHW programs, caregiver education, and other variables described in this paper.

Conclusions
Overall, it appears that structural factors (e.g. vaccine availability at clinics) play a much larger role in determining a child's vaccination status than behavioral factors (e.g. caregiver attitudes about vaccines) in this study population. To improve vaccination rates among infants in Limpopo Province, it is important to prevent vaccine stock-outs, with a particular emphasis on PCV. For this reason, the supply chain of PCV in South Africa should be closely studied in order to identify and address the root causes of PCV stock-outs, with a goal of preventing future stock-outs in those regions of the country with relatively low coverage, like Mopani District. Additionally, given the positive association between caregiver education and vaccination coverage, expanded educational opportunities for women and girls in Limpopo Province may augment vaccination coverage in Limpopo Province. Finally, given the strong positive association between vaccination coverage and recent CHW visits, expanded support for CHW programs, including home visits to recent mothers, may be another means for promoting vaccination among infants in Limpopo Province. However, given the descriptive nature of this crosssectional study, it is difficult to prove causation between any of these associations, and further study of this subject is warranted before policy prescriptions can be confidently made.