Introduction

In 2013 approximately 6.2 million children under the age of five died worldwide, and 3 million of these deaths occurred in Sub-Saharan Africa (SSA) [1]. In 2009, the World Health Organization (WHO) estimated that if global vaccine coverage increased to 90% by 2015, then approximately two million deaths of children under the age of five would be prevented [2]. In the Sub-Saharan African country Uganda, vaccine coverage rates remain well below the WHO goal of 90%, with 82% of children receiving the measles vaccine and 78% completing the three dose series of pentavalent vaccine providing protection against diphtheria, tetanus, pertussis, hepatitis B, and Haemophilis influenza type B (DPT-HB-Hib) in 2013 [3]. One recent study demonstrated that the western region of Uganda, where this study was conducted, has the lowest rate of complete childhood vaccination in the country [4]. Immunizations are a key strategy for reducing the prevalence of infectious diseases, and especially in under-resourced areas, immunizations are a highly cost-effective foundation for developing health systems to invest in [5].

In 2008, the WHO Strategic Advisory Group of Experts on Immunization called for increased information about the factors leading to non-vaccination and under-vaccination of children in order to develop strategies to improve the uptake of childhood immunizations [6]. In rural areas of developing countries, there has been relatively little research into parents’ knowledge and attitudes towards childhood immunizations (KATCI) [7]. Surveying KATCI is an important first step towards understanding the factors that influence vaccine non-acceptance in a particular setting. However, to develop strategies that will improve vaccination rates, the relationship between KATCI and whether they actually have their children adequately vaccinated must be investigated. Community-based sampling, as opposed to, for example, surveying caregivers at a healthcare facility, is particularly important in this context because it ensures broad recruitment inclusive of those most at-risk for under-vaccination [8, 9].

In this study, we aim to first determine basic KATCI by women of childbearing age living in Sheema District, Uganda and to demonstrate how these maternal KATCI correlate with the full, on-time vaccination status of the children of these women.

Methods

Study sample and design

Within the purposively selected villages, 1000 women were surveyed for the cross-sectional study. A needed sample size of 969, which was rounded to 1000, was calculated for two-sided testing of differences in proportions as little as 10% for relatively common factors of interest (overall prevalence = 50%), α = 0.05, and 1-β = 0.80 with a 15% level of non-participation. To allow for village- and region-level analyses with modest statistical power, we set out to recruit a random sample of 100 eligible women from 10 different villages. However, once in the field, we found fewer than 100 eligible women in some villages; thus, women from adjacent villages were approached for the study. No more than one woman per household was recruited.

For each of the households approached, it was first determined which household members were eligible for participation. Eligible participants were female, age 15 to 50 years, verbally confirmed that they had slept at the house the night before, and were willing to consent to the study procedures. If only one person met the eligibility criteria, that individual was asked to participate in the study. If more than one person from the household met the eligibility criteria, then one of them was randomly selected and asked to participate in the study. Of those enrolled in the study, 97.4% were surveyed during the initial visit to the household. Others were not available during the initial visit so interviewers had to return to the household a second (1.8%) or third (0.8%) time to conduct the interview with the selected individual. When recruitment was unsuccessful, either due to resident refusal or three unsuccessful attempts to meet with the selected individual, the next household along the road was approached.

The 1000 women enrolled in the study were interviewed using a standardized questionnaire that collected socio-demographic data and their KATCI. To gauge knowledge, women were surveyed on their understanding of the purpose of vaccinations and the process and timing of immunizing children. To analyze knowledge of vaccine preventable diseases (VPDs), women were asked whether certain diseases are vaccine preventable. Three categories of diseases were asked about:

1. Diseases preventable by WHO recommended vaccines in Uganda at the time of this study (WHO VPDs): polio, hepatitis, whooping cough, diphtheria, tetanus, measles, and meningitis.
2. Other “correct” vaccine preventable diseases (non-WHO VPDs): pneumonia, cancer, diarrhea, and yellow fever.
3. Diseases not preventable with vaccines: HIV and malaria.

To quantify attitudes towards childhood immunizations, the questionnaire covered the importance of immunizations and reasons for low vaccination coverage in these communities.

Vaccination data and date of birth was obtained for children between the ages of one month and five years. The Uganda Ministry of Health provides health cards to caregivers of every child. These cards contain records of children’s birth history, vaccination history, and other pertinent past medical history. Vaccination data were only obtained from vaccination cards sighted during the household visit (Fig 1). When participants included in the study had more than one child within the age range and with a vaccination card available, only vaccination data from the single child closest to one year of age was collected.

Download:

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

Fig 1. Derivation of the sample of children and mothers.

https://doi.org/10.1371/journal.pone.0150131.g001

Results

Sample characteristics

Of the total 1000 women of childbearing age interviewed, 302 had a child between one month and five years of age and a vaccination card for this child at the time of the interview (U5+Card). This group of mother-child pairs was then analyzed for factors associated with full vaccination status. Compared to the entire sample, the women in the U5+Card group had lower representation in the wealthiest quartile, were generally younger, were more likely to be in a union, and were more likely to have slept under a bed net the previous night. Other characteristics of sampled women, as well as the age of children eligible for study inclusion based on age, are shown in Table 1.

Download:

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

Table 1. Socio-demographic characteristics of all women in the study sample and those with a child between one month and 5 years old with a vaccination card (U5+Card).

https://doi.org/10.1371/journal.pone.0150131.t001

Knowledge of childhood immunizations

Table 2 shows various aspects of knowledge of childhood immunizations by all women surveyed and women in the U5+Card group. Compared to the larger group, those in the U5+Card group demonstrated similar knowledge other than being more likely to understand how often an infant needs to be vaccinated (71.2% versus 63.9%, P = 0.0194). None of these other measures of vaccine knowledge were significantly different between the U5+Card subgroup and total sample. The majority of women were able to state that childhood immunizations protect children from diseases (93.5%). Many also thought that they strengthen or improve children’s health (32.8) or promote children’s growth (26.3%). The most common VPDs identified were polio (81.3%) and measles (77.5%). 8.6% of women thought that malaria is vaccine preventable. A majority of women correctly stated how often an infant needs to be vaccinated (63.9%) or the nearest facility for vaccinations (88.9%).

Download:

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

Table 2. Knowledge of childhood immunizations and perceived probability that next child will receive required immunizations by all women in the study sample and those with a child between one month and 5 years age with a vaccination card (U5+Card).

https://doi.org/10.1371/journal.pone.0150131.t002

Fig 2 illustrates the distribution of women identifying certain diseases as vaccine preventable. The women correctly identified a mean of 2.98 (SD = 1.39) out of seven recieveds. Women reporting more WHO VPDs were more likely to mention non-WHO VPDs (r = 0.804). There was no correlation between mentioning malaria and/or HIV as vaccine preventable and number WHO VPDs identified.

Download:

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

Fig 2. Number of diseases identified correctly and incorrectly as vaccine preventable by women surveyed (N = 1000).

https://doi.org/10.1371/journal.pone.0150131.g002

Attitudes towards childhood immunizations

Women endorsing that they wanted to have another child within the next five years were asked to state the probability that child would be fully immunized. As shown in Table 2, the mean probability reported was 0.75 (SD = 0.31). When asked whether it is “very,” “somewhat,” or “not important” for infants to receive vaccinations, 95.7% stated that it is “very important,” 3.9% stated “somewhat important,” and 0.3% stated “not important.” Table 3 shows coded free text of the women’s responses to the question “Why do you think some mothers in your community do not have their children vaccinated at proper times?” The two most common response categories, “fearful of side effects” and “ignorance/disinterest/laziness,” were both over twice as prevalent (44% each) as the next most prevalent response. Reasons related to vaccine supply, “vaccine shortages” and “crowds or long waits,” were only mentioned by 10% of respondents. Responses of women in the U5+Card group were very similar to the distribution of responses for the entire sample.

Download:

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

Table 3. Reasons why parents in their community may not have their children fully vaccinated as reported by all women surveyed (N = 1000) and only those with a child between the ages of one month and five years with a vaccination card (U5+Card, N = 302).

https://doi.org/10.1371/journal.pone.0150131.t003

Factors associated with full vaccination status

Of the 302 children whose vaccination cards were analyzed for this study, 267 (88%) were vaccinated in a timely manner with all vaccines they were eligible for based upon their age. The 35 under-vaccinated children most frequently missed the last polio dose (48%) or measles (43%), as shown in Table 4.

Download:

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

Table 4. Frequency of specific vaccinations missed by children between the ages of one month and five years that had vaccination history documented but were not fully up to date.

https://doi.org/10.1371/journal.pone.0150131.t004

Table 5 explores factors associated with full vaccination status. Women that were able to correctly state that childhood immunizations protect children from diseases were more likely to have their child fully immunized compared to women that were not able to state this benefit of immunizations (PR = 1.354, CI: 1.018, 1.802). Other aspects of vaccine knowledge, including knowledge of specific VPDs and understanding how often and where to get a child vaccinated, were not significantly associated with full vaccination status. The two measures of attitudes towards childhood immunizations, mother’s perceived probability that next child will be fully immunized and the various categories of reasons stated for low vaccination rates in the community, also were not significantly associated with full vaccination status of the child.

Download:

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

Table 5. Bivariate analysis of factors associated with full vaccination among children between one month and 5 years old with vaccination cards.

https://doi.org/10.1371/journal.pone.0150131.t005

The prevalence of children fully vaccinated decreased as reported distance traveled to site of vaccination increased (PR = 0.975, CI: 0.954, 0.996). When stratified by type of travel, this association increased in the group that reported that they walk to vaccination sites (PR = 0.968, CI: 0.940, 0.996). There was no association in the groups that reported that they use public transport, bicycle (n = 11), or private vehicle (n = 1). Socio-demographic covariates including wealth of the household, age of the mother or child, number of people in the household, maternal marital status, maternal desire for more children, and maternal parity were not associated with full vaccination status.

Discussion

This study describes knowledge and attitudes towards childhood immunizations by parous women living in rural western Uganda. In addition, the study describes how these attitudes and knowledge associate with the vaccination status of the children of the women surveyed.

When used to predict factors influencing caregivers’ demand for childhood immunizations, the health belief model predicts that important factors include perception of a child’s susceptibility to VPDs, perception of severity of disease caused by VPDs, impression that immunizations are effective and beneficial, and perception of adequate access to vaccinations [8]. An understanding of the degree to which vaccines are acceptable in communities and the reasons behind any hesitancy from parents to vaccinate their children is important for the success of immunization programs [16]. Studies in SSA analyzing parental characteristics associated with the vaccination status of their respective children have largely focused on other socio-demographic factors. Variables such as parental education [4, 17–23], mothers’ age [18, 19], household income or wealth [4, 11, 17, 19, 22, 24], family size or mothers’ parity [4, 17–19], religion of caregivers [25, 26], and location of labor and delivery [23] are well documented as correlates with the vaccination status of a child. We are only aware of four published studies conducted in SSA in the past 15 years that assessed KATCI [18, 27–29]. Only one of these studies analyzed how maternal attitudes towards immunizations correlate with the vaccination status of a child by undergoing a limited bivariate analysis of mother’s negative attitudes towards local healthcare facilities providing vaccinations and child’s measles vaccination status [18].

In this study, the under-vaccinated group was significantly less likely to know that the main purpose of immunizations is protection from diseases. This is complemented by respondents frequently attributing poor vaccination rates in their community to factors in the category of “ignorance/disinterest/laziness.” VPDs are typically an unseen threat, and the utility of vaccinating children isn’t immediately apparent. When caregivers have basic insight into the larger benefits of vaccinating their children, they may place higher priority on this task amidst multiple other competing interests for their time [8].

Women in this study were asked why parents in their community may not have their children vaccinated, and the two most common responses were “fearful of side effects” and “ignorance/disinterest/laziness.” Two other studies have analyzed caregivers’ perceived reasons for under-vaccination by directly asking the caregivers of under-vaccinated children. Mohamud et al. reported that a population of caregivers with under-vaccinated children in Jigjiga District, Ethiopia most commonly stated “busyness due to work load” as the main reason for not fully vaccinating their respective children [23]. Oria et al. reported that a rural population in western Kenya and an informal settlement in Nairobi most commonly stated “child was sick during vaccination period” and “parent was too busy,” respectively, as the main reason for not vaccinating their children for influenza [30]. Responses similar to “fearful of side effects” and “ignorance/disinterest/laziness” were much less common in both of these surveys. These discrepancies may reflect true differences in barriers to vaccinating children between these different communities. However, the differences may also be attributable to the source of information; this study asked mother’s to generalize about other caregivers in their community while these other two studies asked the caregivers to provide reasons for their own lack of action.

A deeper understanding of the specific VPDs was not associated with the vaccination status of children. These findings are consistent with the majority of evidence from low-income countries, which generally shows that in-depth understanding of vaccinations by caregivers is not an important factor for high immunization coverage [31]. Aspects of caregivers’ KATCI that are thought to correlate with vaccination status of children in low-income countries include positive attitudes towards immunizations and practical knowledge of immunizations such as when to go, how often, and where to go to have a child immunized [31]. We found positive attitudes and practical knowledge to be widespread, but within the constraints of our research methodology, these were not correlates with vaccination status. It is noteworthy that Muhwezi et al. demonstrated in Uganda that some aspects of specific knowledge of cervical cancer and the human papilloma virus (HPV) vaccine correlate with caregivers’ willingness to allow their male children to receive HPV vaccines. Caregivers aware of cervical cancer, aware of a vaccine available to prevent cervical cancer, and aware that HPV causes genital warts were significantly more likely to be willing to have their male children receive the HPV vaccines [32].

Although this group of women most commonly identified polio and measles as VPDs, the 35 under-vaccinated children most commonly missed the last dose of the polio and the measles vaccine. This provides further evidence that knowledge of specific vaccines is poorly associated with their uptake. Administration of BCG and polio at birth is required for any registered maternity health facility in Uganda. Assuming a mother delivers at one of these maternity health facilities, it is possible that factors such as demand and access are relatively less important for receiving these two vaccines. Along these lines, under-vaccinated children in this study were least likely to miss polio at birth, but surprisingly, twice as many did not receive BCG on time.

Many studies in low-income countries have also demonstrated that, particularly in rural areas, long travel distance to vaccination points can be a barrier to immunization uptake [31]. The mothers of children in the under-vaccinated group of this rural population had significantly longer perceived travel distance to the vaccination facility, especially if they walk rather than rely on other modes of transportation. However, less than one-fifth of all respondents reported travel or financial problems as a reason for poor vaccination rates in their community.

In our sample, 88% of children had full vaccination status. Recent estimates from the Uganda Ministry of Health show significantly lower coverage for individual vaccines in Sheema District: 70.3% for BCG, 59.2% for measles, and 66.8% for the final dose of polio [10]. To a lesser extent, the national estimates for individual vaccines in Uganda reported by the WHO and listed in the Background are also lower [3]. These discrepancies are surprising given that we used a more stringent definition of full vaccination status by also requiring timeliness of vaccination. Limiting our analysis only to mother-child pairs with vaccination cards available was the most likely reason for a markedly higher rate of full vaccination status in the population we sampled.

Rwashana et al. developed a model of factors affecting demand for immunization in Uganda using a dynamic synthesis methodology, which combines system dynamics and case study research methods [33]. Their model identifies three main factors influencing demand for immunization in Uganda: level of immunization awareness, mothers’ availability, and level of trust in the health system. As described above, we found level of immunization awareness to correlate with immunization uptake. Many of the reasons reported by these women for low vaccination rates in their communities do coincide with the later two main factors in this model. Future efforts are needed to elucidate the full impact of disinterest or lack of engagement in the vaccination process and concerns about the side effects of vaccinations.

Access to mobile phones for mothers was not significantly different between those with fully and under-vaccinated children. Mobile phone access has been increasing dramatically in rural areas of developing countries such as this over the past decade, and mobile phone-based interventions for improving vaccination coverage in populations at risk for under-vaccination are quickly becoming more feasible and efficacious [34–36]. These interventions may be especially useful in this population where the under-vaccinated group has relatively good mobile phone access.

Knowledge and attitudes towards childhood immunizations vary greatly in different settings [7, 37]. This study was conducted in rural communities populated mostly by subsistence farmers in western Uganda. Although there is much ethnic and socioeconomic diversity in Uganda, it is estimated that 85% of the country’s population lives in rural areas. The results of this study may have implications for vaccination programs in similar settings. However, considering the dearth of information available and high variability in populations that have been studied, we believe basic operational research that elucidates local knowledge and attitudes should be scaled up and considered essential to every vaccination campaign or program. This will help identify root causes of poor demand for vaccinations and allow for effective decisions and policy tailored to fit the needs of communities. This study provides a framework for how this type of operational research can be conducted in developing parts of the world.

Our findings should be interpreted with several limitations of the study in mind. The effect of confounding cannot be underestimated given that we relied on bivariate analysis to determine factors associated with full vaccination coverage. Mother-child pairs were only included in the bivariate analysis if a vaccination card was available at the time of the interview. With no vaccination card available, 174 mother-child pairs eligible based on all other criteria were not included. This exacerbated the issue of small sample size and inadequate power. A study involving mothers in rural South Africa demonstrated that maternal recall of child vaccination dates has high sensitivity when compared to information extracted from vaccination cards [38]. We did attempt to solicit vaccination dates by maternal recall in the group without vaccination cards available. These data were deemed unreliable due to concern for recall bias and because many of the mothers were not able to recall exact dates of vaccinations.

Our survey did not collect information about women’s educational background and place of delivery for mother-child pairs. Both of these variables have been shown to be associated with the vaccination status of children in similar settings [19, 20]. Not accounting for them while analyzing our primary input variables of interest may be a source of bias. Next, although we demonstrated that the prevalence of fully vaccinated children decreases as travel distance to the vaccination site increases, it is important to recognize that we had each of these participants estimate this distance. We didn’t evaluate the accuracy of these estimations, and the amount of bias with these estimations is uncertain. Finally, we only assessed the KATCI of mothers, who certainly have a large influence on the net demand of a particular child’s vaccination. However, a comprehensive analysis of the connection between the vaccination status of a child and the mindset of those caring for the child also must take into consideration KATCI of fathers and other immediate caregivers.

Conclusions

Most studies analyzing factors influencing caregivers demand for childhood immunizations in rural, resource-limited settings do not focus on caregivers’ KATCI. Our analysis shows that in this rural setting of western Uganda, mothers with a basic understanding of the importance of childhood immunizations were more likely to have timely, full vaccination of their children. Many of these women suggested that poor vaccination rates in their community are due to caregivers’ fear of side effects and disinterest or ignorance towards vaccinations. Prospective, larger scale analyses are needed to delineate the community-specific influence caregivers’ KATCI has on children’s vaccination status. This will allow for the development of more effective interventions and policy to improve vaccination coverage in developing countries.

Acknowledgments

This study was funded by a grant from the University of Wisconsin-Madison Global Health Institute. We are grateful to the staff team at MUST Department of Community Health. We also thank the data collectors for their hard work and the study communities for agreeing to participate.